The Turquoise-Chalcosiderite Cu(Al,Fe3+)

American Mineralogist, Volume 96, pages 1433–1442, 2011

3+ The turquoise-chalcosiderite Cu(Al,Fe )6(PO4)4(OH)8·4H2O solid-solution series: A Mössbauer spectroscopy, XRD, EMPA, and FTIR study

Yassir a. abdu,1,* sharon K. hull,2 Mostafa faYeK,1 and franK C. hawthorne1

1Department of Geological Sciences, University of Manitoba, Winnipeg, Manitoba R3T 2N2, Canada 2Department of Anthropology, University of Manitoba, Winnipeg, Manitoba R3T 2N2, Canada

abstraCt Eight turquoise samples covering a wide range of compositions in the turquoise-chalcosiderite solid-solution series were analyzed by Mössbauer spectroscopy, X-ray diffraction (XRD), electron microprobe analysis (EMPA), and Fourier transform infrared (FTIR) spectroscopy. Two of the turquoise samples display evidence of alteration from weathering processes. The unit formulas were calculated on the basis of 24 (O,OH) anions and 11 cations using the results of EMPA, assuming all Fe as Fe3+, as confirmed by Mössbauer spectroscopy. The altered turquoise samples show deficiencies in both cations and anion groups, indicated by EMPA, but they preserve the crystal structure of turquoise, as verified by XRD. They also show large amounts of Si and Ca in their microprobe data, due to the presence of kaolinite and Ca carbonate, respectively, which are identified by FTIR spectroscopy. The isomorphous substitution of Fe3+ for Al in the turquoise structure broadens and shifts the IR bands to lower frequencies, in particular the OH-stretching bands. The Mössbauer spectra, collected at room temperature, are fitted with two generalized Fe3+ sites, using a Voigt-based quadrupole-splitting distribution method, which are assigned to the M3 (smaller quadrupole splitting) on the one hand and M1 and M2 octahedral sites on the other hand. The Fe3+ distribution over the M3 and M1,2 sites, calculated from the Mössbauer relative areas and EMPA, indicates that Fe3+ prefers the larger M3 octahedron in the turquoise-chalcosiderite solid-solution series. Keywords: Turquoise, chalcosiderite, alteration, Castillian Mine, Tyron Mine, Cornwall, electron microprobe analysis, FTIR, Mössbauer spectroscopy

introduCtion (EMPA), FTIR and Mössbauer spectroscopy, to (1) learn more Turquoise is common in the American Southwest, where about turquoise alteration and (2) study the distribution of Fe most of its value is associated with art (e.g., jewelry) and Na- between the metal sites. tive American culture. There has been renewed interest in this baCKground complex mineral due to a recently developed technique that can identify the provenance regions of turquoise artifacts using the Turquoise is a hydrated copper aluminum phosphate and isotopic ratios of hydrogen and copper (Hull et al. 2008). This belongs to a group of minerals, the turquoise group, consisting approach allows archaeologists to reconstruct and investigate of at least six isostructural end-members (Table 1; Foord and pre-Colombian turquoise trade structures, and gemologists to Taggart 1998). The general formula for the turquoise group may 2+ 2+ authenticate the source of turquoise (e.g., Lone Mountain mine be written as A0–1B6(PO4)4(OH)8⋅4H2O with Cu or Fe as the 3+ 3+ in Nevada vs. Sleeping Beauty mine in Arizona). most common constituents at the A position and Al and Fe at 2+ 2+ When affected by near-surface conditions and extended the B position. However, Ca or Zn can occur at the A position exposure to sunlight and meteoric (rain) water, turquoise in some of the more rare members of the turquoise group. The weathers to chalky white clay minerals. This alteration affects range in chemical composition (i.e., different concentrations of identification of the provenance regions of archaeologically Cu, Fe, and Al) of turquoise (sensu lato) produces a wide range recovered turquoise artifacts (Hull et al. 2008), the mineral- of colors (Table 1). Foord and Taggart (1998) suggest that differ- ogical properties of turquoise, and the value of turquoise as a ences in the Cu and Fe ratio are responsible for the differences semi-precious gemstone. However, this alteration process is in color of the samples: blue turquoise has Cu at the A position poorly understood. Therefore, to improve the characterization and Al at the B position, whereas green turquoise (chalcosiderite) 3+ of turquoise provenance for archaeological and gemological largely contains Fe at the B position. purposes, it is important to understand the alteration processes The crystal structure of turquoise is triclinic, space group that affect turquoise. P1, and is illustrated in Figure 1. Different site nomenclatures In this work, we characterize turquoise samples covering have been used in many of the structural and spectroscopic a wide range of compositions in the turquoise-chalcosiderite studies of the minerals of the turquoise group, depending on series by X-ray diffraction (XRD), electron microprobe analysis the chemical compositions under study. Here we use a more general site nomenclature that avoids phrases as for example * E-mail: [email protected] “Fe3+ occupying an Al site”. There are four sites that are octahe-

0003-004X/11/0010–1433$05.00/DOI: 10.2138/am.2011.3658 1433 1434 ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES

Table 1. Minerals of the turquoise group Mineral Chemical formula Color Luster Hardness 2+ Aheylite Fe Al6(PO4)4(OH)8⋅4(H2O) pale blue, green and blue-green vitreous, dull 5–5.5 3+ Chalcosiderite CuFe 6(PO4)4(OH)8⋅4(H2O) apple green, darkgreen vitreous, glassy 4.5 “Coeruleolactite” (Ca,Cu)Al6(PO4)4(OH)8⋅4–5(H2O) milk white, light blue vitreous, waxy 5 Faustite (Zn,Cu)Al6(PO4)4(OH)8⋅4(H2O) apple green chalky, earthy, dull 5.5 Planerite Al6(PO4)2(PO3OH)2(OH)8⋅4(H2O) white, olive green, pale blue, green and blue-green vitreous, dull 5 Turquoise CuAl6(PO4)4(OH)8⋅4(H2O) pale green, blue-green, turquoise blue waxy 5–6 Note: Mineral and chemical formulas from Foord and Taggart (1998).

(H2O) (OH) Al sites (i.e., M sites in our notation). However, Giuseppetti et al. T1 (1989) refined the crystal structure of an Al-bearing chalcosider- 3+ M2 M1 ite (Al = 0.54 apfu) and found that Fe is partly ordered at the Fe1 T2 site (corresponds to Al3 site in the work of Cid-Dresdner 1965). The X Al1 and Al2 sites were labeled by Giuseppetti et al. (1989) as Fe2A M3 and Fe2B, respectively. Giuseppetti et al. (1989) also reanalyzed the powder X-ray diffraction data of Cid-Dresdner and Villarroel (1972) for rashleighite and cautiously suggested similar behavior of Fe3+ in intermediate compositions of the series, where it is very difficult a b to find single crystals suitable for X-ray diffraction. 57Fe Mössbauer spectroscopy, due to its ability to probe the c local environment of the Fe nucleus regardless of the degree of sample crystallinity, can be used to verify the Fe3+ site preference in the turquoise-chalcosiderite series. To our knowledge, the work figure 1. The crystal structure of turquoise; M1 and M2 octahedra are shown in yellow, M3 octahedra are shown in pink, X octahedra are done on turquoise that involves Mössbauer spectroscopy is very shown in green, and T tetrahedra are shown in blue; hydrogen bonds are scarce (Belyaev and Ievlev 1989; Sklavounos et al. 1992), and omitted for clarity. (OH) and (H2O) groups are shown as red and blue the technique has never been used to address the above issue circles, respectively. in turquoise. Other Mössbauer studies of hydrated phosphate minerals are reported by Amthauer and Rossman (1984), Da drally coordinated by O2– and (OH)− anions and two T sites that Costa et al. (2005), and De Resende et al. (2008). are tetrahedrally coordinated by O2– anions, and are occupied Fourier-transform infrared (FTIR) spectroscopy is a valuable by P. The octahedrally coordinated sites may be divided into technique in studying hydrated minerals. It is very sensitive to two groups, the M1–M3 sites that are occupied dominantly the hydroxyl (OH) group and can easily distinguish between OH by small trivalent cations, and the X site that is occupied by in the structure and H2O molecules. The technique can be used medium-sized divalent cations (primarily Cu2+, Fe2+, and Zn). to supplement X-ray diffraction and chemical analysis, as FTIR

Thus, the structural formula for turquoise is X (M12M22M32)Σ=6 spectra can be easily obtained for crystalline as well as poorly

(PO4)4(OH)8·4H2O. In the structure of turquoise (Fig. 1), pairs crystalline materials. Recently, Reddy et al. (2006) studied in of edge-sharing M1 and M2 octahedra are linked by sharing detail two Fe-bearing turquoise samples from Arizona (U.S.A.) corners with pairs of T tetrahedra to form chains that extend in and the Kouroudaiko mine (Senegal), with Fe2O3 contents of 1.94 the b direction. These chains are linked in the a and c directions and 3.19 wt%, respectively, by FTIR spectroscopy and assigned by sharing corners with M3 octahedra, and further linkage is the different vibrational bands for both the hydroxyl (OH) and 3– provided by X octahedra that share edges with the M1 and M2 the phosphate (PO4) groups. octahedra. The heteropolyhedral framework is strengthened by a network of hydrogen bonds involving the (OH) groups in the Materials and experiMental Methods formula. Of particular importance in Figure 1 are the similarities The provenances of the turquoise samples used in this study are given in and differences exhibited by the M1–M3 octahedra. The M1 and Table 2. The samples were characterized by powder XRD using a Philips (PANalyti- cal) PW1710 automated diffractometer with CuKα radiation. The PW1710 system M2 octahedra are very similar in their general stereochemistry consists of a 4kW sealed-tube type X-ray generator in a housing with a centrally and their linkage to adjacent cation polyhedra is identical. On located tube tower and vertical goniometer with Bragg-Brentano (θ–2θ) geometry. the other hand, the local stereochemistry of the M3 octahedron The goniometer is equipped with a graphite monochromator in the diffracted beam is very different from that of M1 and M2. It does not share any and a scintillation detector. The microprocessor is operated remotely from a PC using MDI DataScan software and data are processed using MDI Jade+ software. edges with the M1 and M2 octahedra and shares a corner with Backscattered electron images and electron microprobe analysis were done the X octahedron (Fig. 1). Thus similar behavior is expected from constituents occupying the M1 and M2 octahedra and dif- Table 2. Sample provenance ferent behavior from constituents occupying the M3 octahedron. Sample(s) Provenance 3+ Substitution of Al by Fe results in the solid-solution se- CS-1, Cas-1, Cas-4a, Cas-4c Castillian Mine, Cerrillos Hills, 3+ ries turquoise-chalcosiderite: Cu(Al,Fe )6(PO4)4(OH)8·4H2O. New Mexico, U.S.A. Cid-Dresdner and Villarroel (1972) studied rashleighite, an Ty-1 Tyron Mine, New Mexico, U.S.A. GT-1 Green Tree, Nevada, U.S.A. intermediate member of the series, by powder X-ray diffraction, Ry-1 Royston, Nevada, U.S.A. and concluded that Fe3+ is statistically distributed over the three Chalcosiderite Cornwall, England ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES 1435 using a CAMECA SX-100. Chemical analysis were obtained using wavelength- collected within the mine pit supports the assumption that the dispersion mode with the following conditions: excitation voltage: 15 kV; specimen most exposed samples were the most altered, because these current: 10 nA; beam size: 10 µm. Mean chemical compositions for all samples are given in Table 3, together with the unit formulas normalized to 24 (O, OH) were much lighter in color and could be easily crumbled with anions and 11 cations. As can be seen from Table 3, the chemical compositions of the slightest pressure (Figs. 2a and 2b). The samples were col- the turquoise samples cover a wide range of the turquoise-chalcosiderite series. lected from fractures and veinlets where turquoise occurs along the The Al-bearing chalcosiderite used in this study is from the same locality as that wall of the 25 m deep mine pit. The least-altered Castillian sample, studied by Giuseppetti et al. (1989), Cornwall, England, and its Fe and Al contents Cas-1, was collected from the bottom of the pit, Cas-4a (altered) obtained by microprobe analysis (Table 3) are similar to those determined by X-ray crystal-structure refinement (Giuseppetti et al. 1989). was collected approximately 13 m below the ground surface, and FTIR spectra were collected using a Bruker Tensor 27 FTIR spectrometer Cas-4c (altered) was collected from a vein that ran vertically through equipped with a KBr beam splitter and a DLATGS detector. Spectra over the the mine pit located near Cas-4a. SEM images of the surface of –1 range 4000–400 cm were obtained by averaging 100 scans with a resolution of the unaltered turquoise sample (Cas-1) show a blocky, crystalline 4 cm–1. Powdered samples were prepared as KBr pellets, with a sample/KBr ratio of ~1%. Base-line correction was done using the OPUS spectroscopic software texture (Fig. 3a), whereas altered turquoise (Cas-4a) has a platy (Bruker Optic GmbH). surface (Fig. 3b). Mössbauer spectroscopy measurements were done at room temperature (RT) using a 57Co(Rh) point source. Mössbauer absorbers were prepared by mixing the sample with powdered sugar and loading the mixture into a sample holder containing ~5 mg Fe/cm2. The spectrometer was calibrated with the RT spec- a) trum of α-Fe in the velocity range of ±4 mm/s. The spectra were analyzed by a Voigt-based quadrupole-splitting-distribution (QSD) method (Rancourt and Ping 1991) implemented in the Recoil software. To account for absorber-thickness and instrumental broadening, the Lorentzian linewidth of the elemental doublets of the QSD was allowed to vary during the spectrum fitting procedure (Rancourt 1994).

results and disCussion Alteration Three samples were collected from the Castillian Mine pit (Cas-1, Cas-4a, Cas-4c) and one sample (CS-1) was provided by Douglas Magnus, owner of the Castillian turquoise mine, New Mexico. These samples showed variable alteration. On visual inspection, the variation of hardness and colors of the samples

Table 3. Average chemical composition (wt%) and unit formula Cas-1 (apfu), based on 24 (O,OH) and 11 cations, for turquoise GT-1 Ry-1 Ty-1 CS-1 Cas-4a Cas-4c Cas-1 Chalco- siderite no.* 10 5 9 10 10 10 10 10 P2O5 34.43 35.58 31.45 32.78 21.96 26.24 31.07 28.50 b) SiO2 0.02 0.38 1.07 0.03 18.13 5.82 0.02 0.06 SO3 0.17 0.14 0.21 0.56 0.40 4.56 1.21 n.a. Al2O3 35.68 35.71 26.45 26.73 14.45 20.46 18.04 2.56 Fe2O3 2.07 2.50 14.38 14.53 16.87 16.45 25.49 44.07 Cr2O3 0.20 n.a. 0.02 0.02 0.02 0.01 0.02 0.02 CuO 8.96 8.82 8.96 7.68 5.99 6.66 7.23 8.11 ZnO 0.18 n.a. 0.04 0.08 0.07 0.07 0.01 0.08

TiO2 0.02 n.a. 0.01 n.d. n.d. n.d. 0.09 n.d. MgO 0.01 n.a. 0.02 0.01 0.44 0.21 n.d. n.d. CaO 0.22 0.07 0.07 0.12 8.89 3.70 0.01 0.03

K2O 0.06 n.a. 0.08 0.11 0.11 0.07 0.08 0.05 H2O† 18.00 16.84 17.27 17.46 12.74 16.64 16.96 16.49 Total 100.02 100.04 100.03 100.11 100.07 100.89 100.23 99.97 P 3.99(3) 4.08(2) 3.81(4) 3.97(6) 2.50(39) 3.16(11) 3.95(4) 3.98(5) Si 0.00 0.05 0.15(2) 0.00 2.4(8) 0.83(18) 0.00 0.01 S 0.02 0.01 0.02 0.06 0.04 0.49(19) 0.14(1) n.a. Al 5.76(3) 5.69(3) 4.46(5) 4.51(14) 2.29(40) 3.43(20) 3.19(14) 0.50(11) Fe 0.21(1) 0.25(1) 1.55(2) 1.57(11) 1.71(26) 1.76(10) 2.88(15) 5.47(6) Cr 0.02 n.a. 0.00 0.00 0.00 0.00 0.00 0.00 Ti 0.00 n.a. 0.00 n.d. n.d. n.d. 0.01 n.d. Cu 0.93(1) 0.90(1) 0.97(2) 0.83(3) 0.61(9) 0.71(5) 0.82(3) 1.01(2) Cas-4c Zn 0.02 n.a. 0.00 0.01 0.01 0.01 0.00 0.01 Mg 0.00 n.a. 0.00 0.00 0.09(3) 0.04 n.d. n.d. Ca 0.03 0.01 0.01 0.02 1.3(12) 0.56(25) 0.00 0.01 K 0.01 n.a. 0.01 0.02 0.02 0.01 0.02 0.01 figure 2. Images of turquoise hand specimens. (a) Dark green, OH‡ 16.4(3) 15.2(2) 16.5(5) 16.7(6) 11.4(3) 15.8(5) 17.0(5) 18.1(3) least altered Castillian sample, Cas-1, collected from the bottom of the Notes: n.a. = not analyzed, n.d. = not detected. * Number of spots. Castillian mine pit; (b) Cas-4c (altered) was collected from a vein that † Calculated by difference. ran vertically through the mine pit. Note the lighter green color and ‡ OH + H2O. crumbling nature of the altered turquoise. 1436 ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES

a) a) pluck pits

turq

blocky turq

b) b)

alt turq

Low-Fe

platy

High-Fe

figure 3. SEM images of (a) unaltered turquoise (sample Cas-1) figure 4. Backscattered electron images of (a) Cas-1, unaltered with a crystalline blocky morphology, and (b) altered turquoise (Cas-4a) turquoise, showing a relatively homogenous reflectance and pluck pits with a platy surface. due to polishing, and (b) Cas-4a, altered turquoise, with a much more porous and fibrous surface consisting of patches with low and high Fe. Detailed petrography of the altered and unaltered samples shows distinct textural differences (Fig. 4). Backscattered a negative correlation between P2O5 and (SiO2+SO3) (Fig. 6), electron images (BSE) of unaltered turquoise (Cas-1) show a which may indicate that P is leached from the turquoise structure relatively homogenous texture with a few pluck marks due to and replaced by Si and/or S during the alteration process. The polishing, which appear as voids (Fig. 4a). Altered turquoise samples also contain Ca, the abundance of which correlates with (Cas-4a) shows a highly porous and fibrous texture with rosettes the Cu deficiency (Table 3). It is unlikely that Ca replaces Cu at that are higher in Fe contents than the matrix (Fig. 4b). How- the A site, and the existence of a Ca-dominant A site member of ever, altered turquoise generally has lower Fe content relative the turquoise group “coeruleolactite” is doubted by Foord and to unaltered turquoise (Fig. 5a). Taggart (1998). They further suggested that the A-site deficiency

Figure 5a shows the variation of Al2O3 as a function of Fe2O3 in turquoise, which is accompanied by an excess of H2O, can be 3– for all samples using the microprobe data of all analyzed spots. accounted for by protonation of the (PO4) groups to form plane-

Data for the majority of the turquoise samples lie along the ideal rite [Al6(PO4)2(PO3OH)2(OH)8⋅4(H2O)]. The A-site deficiency line (dashed line in Fig. 5a) connecting the end-member composi- in the altered samples is accompanied by lower H2O contents, tions of the turquoise-chalcosiderite series, ideal turquoise (Al2O3 as well as deficiencies at the B and P sites (Table 3), but the

= 37.60 wt%) and ideal chalcosiderite (Fe2O3 = 48.55 wt%). crystal structure of turquoise is maintained, as verified by XRD However, the data points of the altered turquoise samples (Cas-4a (Fig. 7). This is in agreement with the work of Van Wambeke and Cas-4c) deviate significantly from this line (Fig. 5a) because (1971) on the alteration of some phosphates, where the alteration the B site in the altered samples has variable cation deficiency, process can sometimes cause cation and anion deficiencies, with i.e., Al + Fe <6 apfu (Fig. 5b, Table 3). The altered samples show preservation of the structure. Thus, the presence of Si and Ca in ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES 1437

40 30 Cas-1 Cas-1 35 Cas-4a Cas-4a 25 Cas-4c Cas-4c CS-1 30 CS-1 20 Ry-1 Ry-1 25 Ty-1 Ty-1 (wt%) 3 GT-1 GT-1 15 (wt%) 20 Chalcosiderite 3 + SO 2

O Chalcosiderite 2 Al 15 SiO 10

10 5 (a) 5 (a) 0

0 0 5 10 15 20 25 30 35 40

0 10 20 30 40 50 60 P2O5 (wt%)

Fe2O3 (wt%) 3.0

2.5 Cas-4a 5 2.0

4 1.5 Cas-4c Cas-4c

3 Si+S (apfu) 1.0 Al (apfu) Cas-4a 2 0.5 (b) 1 0.0 (b) 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 0 P (apfu) 0123456 figure 6. (a) (SiO2+SO3) vs. P2O5 of all microprobe data for 3+ Fe (apfu) turquoises; (b) average (Si+S) vs. P (apfu) for turquoises plotted using the data in Table 3. Error bars are smaller than the symbol when not visible. figure 5. (a) Al2O3 vs. Fe2O3 of all microprobe data for turquoises; (b) average Al vs. Fe3+ (apfu) for turquoises plotted from the data in Table 3. Error bars are smaller than the symbol when not visible. The dashed line connects ideal-turquoise to ideal-chalcosdierite. molecules. The P-O stretching vibrations of the phosphate group 3– –1 (PO4) are located between 1200–900 cm and the bands in the the altered samples, as indicated by EMPA (Table 3), could be region below 900 cm–1 are mainly due to coupled motions of the due to impurity phase(s) that escaped detection by XRD, due to frameworks of tetrahedra and octahedra its very low concentration or poor crystallinity. The unit cell of turquoise contains four OH groups and two

Based on EMPA and detailed petrography, the alteration of H2O molecules. Thus, an FTIR spectrum of turquoise in the OH turquoise could be the result of the following reaction region is expected to show four sharp OH-stretching bands from

the OH groups and two broad bands from the H2O molecules. 3+ 3+ + 2– CuFe6(PO4)4(OH)8·4H2O + H2O + 2Al + 4Si + Na + 3SO4 + The FTIR spectra of the GT-1 and Ry-1 samples are very similar, chalcosiderite aqueous and show five bands in the OH-stretching region. For GT-1, three sharp OH-stretching bands occur at 3510, 3467, and 3451 cm–1, 4O = Al Si O (OH) + NaFe3+ (SO ) (OH) + 2(g) 2 4 10 2 3 4 2 6 and two broad bands of the H O molecules are found at 3287 pyrophyllite natrojarosite 2 and 3070 cm–1 (Fig. 8), very similar to the bands reported for 3+ Fe2O3 + CuSO4 + 4H3PO4 + Fe . Arizona turquoise by Reddy et al. (2006). With increasing Fe hematite aqueous content, the OH-stretching bands become broader, the spectra gradually lose resolution, and the bands shift to slightly lower frequencies, which is most evident in the spectrum of chalco- FTIR siderite (Fig. 8). The FTIR spectra of Cas-4a, Cas-4c, and Ty-1 Figure 8 shows the FTIR spectra in the range 4000–400 samples show two extra OH-stretching bands at ∼3698 and 3622 cm–1 for all samples. The bands in the region 4000–2500 cm–1 cm–1, which are more pronounced in the spectrum of Cas-4c (Fig. are due to OH-stretching vibrations of the OH groups and H2O 8). Figure 9 shows an expanded view of the region 3720–3600 molecules in the turquoise structure. The broad band at 1600 cm–1 for the Cas-4c spectrum, where very weak bands at 3675, –1 –1 cm is associated with the H-O-H bending vibrations of the H2O 3668, 3655, and 3648 cm can be observed between the 3698 and 1438 ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES

P-O stretch

200 H-O-H O-H stretch bend

100 Cas-4c Chalcosiderite 5.47

200 2.88 Cas-1

100 1.76

Intensity (a.u.) Cas-4a Cas-4c

C-O 0 stretch

1.71 Cas-4a 200 Absorbance

100 1.57 CS-1 Cas-1

0 1.55 Ty-1 10 20 30 40 50 60

2 figure 7. Powder X-ray diffrcation patterns of the unaltered turquoise (Cas-1) and the altered turquoises (Cas-4a and Cas-4c). All diffraction lines belong to the turquoise structure. 0.25 3467 Ry-1 3451

–1

3622 cm bands. These OH-stretching bands are characteristic 3287 3070 of kaolinite-group minerals (Prost et al. 1989; Johansson et al. 3510 1998; Madejová and Komadel 2001). The Si-O stretching bands of kaolinite, which occur at ~1000–1100 cm–1, are not resolved 0.21 GT-1 due to the strong overlap with the P-O stretching bands of the 3– 4000 3000 2000 1000 (PO4) group in turquoise. However, the Si-O-Si bending band at –1 ∼470 cm , as well as the OH-bending of the inner OH groups at Wavenumber (cm-1) ∼912 cm–1 (Madejová and Komadel 2001), can easily be identi- figure 8. FTIR spectra for turquoises. The numbers on the left side fied in the FTIR spectra of Cas-4c and Cas-4a samples (Fig. 10). are the total Fe content in apfu. The small sharp peak at ~1385 cm–1 is The FTIR spectra of the Cas-4a and Cas-4c samples (Fig. 8) an artifact from the KBr pellet. also show weak bands centered at 1430 and 1460 cm–1, characteristic of the C-O stretching vibrations of the carbonate group 2– (CO3) (Adler and Kerr 1963). The high Ca content indicated No such band was observed in the FTIR spectra of our turquoise by the electron microprobe data suggests that it is a Ca-rich samples, which have variable amounts of Fe3+. Therefore, the carbonate. Therefore, the high Si and Ca contents of the altered 3608 cm–1 band in the FTIR spectrum of the Senegal turquoise turquoise samples are attributed to the presence of kaolinite and is most likely due to an impurity phase. Ca carbonate, respectively, admixed with turquoise. 3+ In the study of Reddy et al. (2006), the FTIR spectrum of the Site preference of Fe turquoise from Arizona (Fe2O3 = 1.94 wt%) showed three OH- Mössbauer spectra. Selected room-temperature Mössbauer stretching bands, similar to those observed in the FTIR spectrum spectra of turquoise samples are shown in Figure 11. They gen- 3+ of the GT-1 sample (Fe2O3 = 2.07 wt%). However, the turquoise erally consist of a broad Fe doublet centered at ∼0.1 and 0.7 from Senegal (Fe2O3 = 3.19 wt%) shows an additional narrow mm/s, and a visible shoulder on the low-velocity peak, indicat- OH-stretching band at 3608 cm–1, which was assigned to the ing the presence of at least two Fe3+ sites. Fe2+, with its high- OH-Fe3+ vibrations in the turquoise structure (Reddy et al. 2006). velocity peak located between 2 and 3 mm/s, was not detected ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES 1439

electric-field gradient. In general, the QS for Fe3+ is expected to increase with increasing distortion of the surrounding coordina- Cas-4c tion polyhedron. The large difference between the quadrupole splittings of Fe3+ (M3) and Fe3+ (M1,2) suggests that these poly- 3622 hedra differ in distortion (Table 4). The distortion parameters for 3674 3668 3653 3648 the M1, M2, and M3 octahedra, calculated for turquoise (Cid- Dresdner 1965) and chalcosiderite (Giuseppetti et al. 1989) are listed in Table 5. Evidently, the M1 and M2 octahedra are more distorted and have similar distortion parameters, while the M3 octahedron is less distorted. Thus, the smaller QS of Fe3+ (M3) Absorbance 3648 compared to that of Fe3+ (M1,2) can be correlated with the smaller

3655 distortion of the M3 octahedron relative to M1 and M2 octahedra. This is in accord with the positive correlation between QS and 3698

3668 3+

3675 polyhedral distortion generally observed in Fe spectra (Patrier et al. 1991). The spectral contributions of Fe3+ at the individual M1 and M2 sites, which are geometrically very similar, are strongly overlapping and cannot be resolved. The Fe3+ site assignment made above can be verified using 3720 3700 3680 3660 3640 3620 3600

-1 single-crystal structure refinement (SREF). However, with the Wavenumber (cm ) exception of the Cornwall chalcosiderite, which was studied figure 9. FTIR spectrum of the Cas-4c sample in the region 3720– 3600 cm–1 showing the kaolinite OH-stretching bands. For comparison, the position of the inner bands of the OH-stretching vibrations of kaolinite reported by Johansson et al. (1998), Figure 1, are shown (dotted lines). in any of the spectra. Consequently, the spectra were fit using a Voigt-based quadrupole-splitting distribution (QSD) to a model having two Fe3+ sites (each with a single Gaussian component). 470 912 The hyperfine parameters for all samples are given in Table 4. The site with the smaller average quadrupole splitting (QS) is assigned to Fe3+ at M3 and that with the larger QS to Fe3+ at the M1 and M2 sites (=M1,2) based on previous work (Sklavounos et al. 1992) and the following discussion. Sklavounos et al. (1992) studied a turquoise sample with an Fe content of 0.13 apfu and Cas-1 fitted its Mössbauer spectrum with two Lorentzian doublets. They tentatively assigned the outer doublet (QS = 1.09 mm/s) to Fe3+ at the Al3 site (M3) and the inner doublet (QS = 0.52 mm/s) 3+ to Fe at both Al1 and Al2 sites (M1 and M2), and concluded Cas-4c that more spectra of turquoises having different Fe contents are Absorbance needed to make more reliable site assignment for Fe3+. The center shift (CS), ~0.41 mm/s for Fe3+ (M3) and ∼0.37 mm/s for Fe3+ (M1,2) is characteristic of Fe3+ in octahedral coordination (Table 4). The QS for Fe3+ arises only from the lattice contribution, as the 3d5 electronic-charge distribution is Cas-4a spherically symmetric and hence has a zero contribution to the

Table 4. Room-temperature Mössbauer parameters for turquoise samples Fe3+ (M3) Fe3+ (M1,2) Sample CS QS A CS QS A (mm/s) (mm/s) (%) (mm/s) (mm/s) (%) CS-1 GT-1 0.406(8) 0.61(9) 65(19) 0.38(2) 1.0(1) 35(19) Ry-1 0.41(1) 0.54(5) 69(18) 0.37(2) 1.1(3) 31(18) Ty-1 0.411(4) 0.55(2) 81(5) 0.38(2) 1.14(8) 19(5) CS-1 0.407(4) 0.54(2) 53(4) 0.372(5) 1.05(4) 47(4) 900 800 700 600 500 400 Cas-4a 0.399(3) 0.53(2) 60(3) 0.370(5) 1.12(3) 40(3) -1 Wavenumber (cm ) Cas-4c 0.407(4) 0.53(2) 54(3) 0.370(4) 1.14(2) 46(3) Cas-1 0.409(6) 0.57(4) 53(5) 0.379(6) 1.10(4) 47(5) Chalcosiderite 0.419(5) 0.48(2) 37(2) 0.381(3) 1.10(1) 63(2) figure 10. FTIR spectra of the Castillian turquoises in the region –1 Notes: CS = Center shift relative to α-Fe at RT, QS = quadrupole splitting, A = below 1000 cm . The vertical dahsed lines mark the positions of the –1 Mössbauer relative area. Standard deviations for the Mössbauer parameters Si-O-Si bending (470 cm ) and OH bending of the inner OH groups were calculated using the covariance matrix of the fit. (912 cm–1) in kaolinite. 1440 ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES

100 100

98 98

96 96 Transmission(%) Transmission(%) 94 94 CS-1 Cas-4a

-2 -1 0123 -2 -1 0123

Velocity (mm/s) Velocity (mm/s)

100 100

98 98

96 96 Transmission (%) Transmission (%)

94 Ty-1 94 Cas-1

-2 -1 0123 -2 -1 0123

Velocity (mm/s) Velocity (mm/s)

100 100

99 98

96 Transmission(%)

Transmission (%) Transmission 97 GT-1 Chalcosiderite 96

-2 -1 0123 -2 -1 0123 Velocity (mm/s) Velocity (mm/s) figure 11. Selected room-temperature Mössbauer spectra for turquoises. Solid-line subspectra: Fe3+ (M3), dashed-line subspectra: Fe3+ (M1,2). Residuals are shown above each spectrum.

by Giuseppetti et al. (1989), the other turquoise samples do Table 5. Distortion parameters of the M-octahedra in turquoise and not contain single crystals suitable for SREF. The SREF data chalcosiderite calculated from the structural data of Cid- Dresdner (1965) and Giuseppetti et al. (1989), respectively on chalcosiderite (Giuseppetti et al. 1989), which to our best Sample Site λ σ2 ∆ V knowledge is the only available SREF data on an Fe-containing Turquoise M1 1.0145 42.21 0.0016 8.98 3+ turquoise, indicate that the Fe occupancies at the M1, M2, and M2 1.0161 42.93 0.0025 9.03 M3 sites are 0.88, 0.88, and 0.97, respectively. This means that M3 1.0022 4.70 0.0006 9.38 3+ Chalcosiderite M1 1.0172 53.47 0.0013 10.44 36% of Fe is located at the M3 site and 64% at the two other M2 1.0161 46.69 0.0020 10.49 sites (M1 and M2), in excellent agreement with our Mössbauer M3 1.0023 4.98 0.0006 10.85 data for chalcosiderite (Table 4). Note: λ and σ2 are the mean octahedral quadratic elongation and octahedral Figure 12 shows the variation of the QS as a function of angle variance, respectively (Robinson et al. 1971); ∆ is the bond-length distortion (Fleet 1976); V is the octahedral volume. Estimated errors are better than composition [expressed as Fe/(Fe+Al)] for all samples. It seems 0.0005 for λ and ∆ and better than 0.05 for σ2 and V. ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES 1441 from the figure that the substitution of Fe3+ for Al at the M sites between the M3 and M1,2 sites. The Fe3+ site occupancies at in the turquoise-chalcosiderite series has no significant effect M3 [XFe(M3)] and M1,2 [XFe(M1,2)] are calculated by multiplying 3+ 3+ 3+ on the QS of Fe . This may indicate that the distortions of the the Fe (M3)/Fetot and Fe (M1,2)/Fetot ratios by the total Fe M-octahedra have not been changed significantly by the sub- obtained by microprobe analysis and dividing by the site(s) stitution, in agreement with the crystal-structure data (Table 5). multiplicity. These occupancies can be related to a distribution 3+ The CS of Fe (M3) is significantly larger than the CS for parameter, KD, defined by Fe3+ (M1,2), Table 4. Sklavounos et al. (1992) attributed this to the larger volume of the M3 octahedron compared to the M1 and KD = XFe(M1,2) [1 – XFe(M3)]/XFe(M3) [1 – XFe(M1,2)], M2 octahedra. A larger volume would result in a decrease in the 3+ s-electron density at the Fe nucleus and hence an increase in the where, KD < 1 indicates preference of Fe for the M3 site, KD > 3+ CS. However, the relative increase in volume between the M1 1 corresponds to ordering of Fe at the M1,2 sites, and KD = 1 (or M2) and M3 octahedra is just ∼5% (Table 5). The octahedral indicates complete disorder. 3+ 3+ volume increase as a result of the substitution of Fe (ionic ra- Fe site occupancies, XFe(M3) and XFe(M1,2), together with 3+ dius ≈ 0.65 Å) for Al (ionic radius ≈ 0.54 Å) from turquoise to the KD values, are listed in Table 6. The distribution of Fe chalcosiderite is ∼16% (Table 5) and yet the CS of Fe3+ at the M between the M3 and M1,2 sites is shown in Figure 13. The sites does not show significant variation with composition (Table solid line in the figure represents complete disorder (i.e., KD 4). Therefore, the difference in the CS between Fe3+ (M3) and = 1), and the dashed curve is the calculated distribution using 3+ Fe (M1,2) cannot be due to the difference in volume between KD = 0.24 (average value, excluding Ty-1 and chalcosiderite). the M3 and M1,2 octahedra. The Fe3+ is bonded to two O, three For comparison, we show on the same plot the data for chalco-

OH, and one H2O at the M1 and M2 sites, and four O and two siderite obtained from structure refinement (Giuseppetti et al. OH at the M3 site (Fig. 1). Molecular orbital calculations have 1989), which agrees with the Mössbauer data. Figure 13 clearly shown that, when OH groups replace some of the O ligands in indicates that Fe3+ cations have preference for the M3 site in the Fe3+ octahedral coordination environment, the covalent in- the turquoise-chalcosiderite solid-solution series. Patrier et al. teraction between Fe3+ and the remaining O ligands is enhanced (1991) observed similar behavior of Fe3+ in epidote, where Fe3+ (Sherman 1985). Because the CS is negatively correlated with orders at the larger octahedron, M3, while the smaller M1 and the degree of covalency for 57Fe, the smaller CS for Fe3+ (M1,2) could be due to a relatively larger degree of covalency of the Table 6. Calculated Fe3+ site occupancies for turquoise samples Fe-O bonds in the M1 and M2 octahedra. Sample Fetot (apfu) XFe(M3) XFe(M1,2) KD 3+ GT-1 0.21(1) 0.07(2) 0.02(1) 0.27(16) Fe distribution over the M1–M3 sites. Assuming equal Ry-1 0.25(1) 0.09(2) 0.02(1) 0.21(11) 3+ 3+ recoilless fractions for Fe (M3) and Fe (M1,2), the Mössbauer Ty-1 1.55(2) 0.63(4) 0.07(2) 0.04(1) 3+ 3+ CS-1 1.57(11) 0.42(5) 0.18(2) 0.30(6) relative areas (Table 4) represent the Fe (M3)/Fetot and Fe 3+ 3+ Cas-4a 1.71(26) 0.51(8) 0.17(3) 0.20(6) (M1,2)/Fetot ratios. As stated above, the Fe (M1) and Fe (M2) Cas-4c 1.76(10) 0.48(4) 0.20(2) 0.27(4) spectral contributions are not resolvable in our room temperature Cas-1 2.88(15) 0.76(8) 0.34(4) 0.16(6) Mössbauer spectra. However, the Fe3+ occupancies at the M1, Chalcosiderite 5.47(6) 1.00(6) 0.86(3) 0 M2, and M3 sites in the Al-bearing chalcosiderite, determined Note: KD = XFe(M1,2) [1 – XFe(M3)]/XFe(M3) [1 – XFe(M1,2)]. by structure refinement, are 0.88, 0.88, and 0.97, respectively, 3+ indicating equal distribution of Fe between the M1 and M2 1.0 sites (Giuseppetti et al. 1989). Assuming similar behavior for the distribution of Fe3+ between the M1 and M2 sites in the other turquoise samples, we can study the Fe3+-Al order-disorder 0.8 Ty-1

1.4 0.6 Fe(M3) X 1.2 0.4

M1,2 1.0 0.2 QS (mm/s) 0.8

0.0 0.6 0.0 0.2 0.4 0.6 0.8 1.0

XFe(M1,2) M3 0.4 figure 13. Fe3+ distribution over the M3 and M1,2 sites. Solid line: 0.0 0.2 0.4 0.6 0.8 1.0 complete disorder, i.e., KD = 1; dashed curve: calculated distribution Fe/(Fe+Al) using KD = 0.24; open circle: data from Giuseppetti et al. (1989) obtained figure 12. Variation of QS with Fe/(Fe+Al) for Fe3+ (M3) and Fe3+ by structure refinement. Error bars are smaller than the symbol when (M1,2). Error bars are smaller than the symbol when not visible. not visible. 1442 ABDU ET AL.: THE TURQUOISE-CHALCOSIDERITE SOLID-SOLUTION SERIES

3+ M2 octahedra are predominately occupied by Al. chalcosiderite, CuFe6 (PO)4(OH)8.4H2O. Neues Jahrbuch für Mineralogie Monatshefte, 227–239. The turquoise sample from the Tyron Mine, Ty-1, shows Hull, S., Fayek, M., Mathien, F.J., Shelley, P., and Roler Durand, K. (2008) A new strong order of Fe3+ at the M3 site compared to turquoises from approach to determining the geological provenance of turquoise artifacts us- the Castillian Mine, Figure 13, which is not obvious, given that ing hydrogen and copper stable isotopes. Journal of Archaeological Science, 35, 1355–1369. Ty-1 and CS-1 samples have similar Fe content (Table 3). This Johansson, U., Holmgren, A., Forsling, W., and Frost, R. (1998) Isotopic exchange difference in Fe3+ ordering may be related to the P-T conditions of kaolinite hydroxyl protons: A diffuse reflectance infrared Fourier transform under which turquoises form, which are currently not well spectroscopy study. Analyst, 123, 641–645. Madejová, J. and Komadel, P. (2001) Base line studies of the clay minerals society established. source clays: Infrared methods. Clays and Clay Minerals, 49, 410–432. Patrier, P., Beaufort, D., Meunier, A., Eymery, J-P., and Petit, S. (1991) Determina- aCKnowledgMents tion of the nonequilibrium ordering state in epidote from ancient geothermal field of Saint Martin: Application of Mössbauer spectroscopy. American We thank Douglas Magnus (Castillian Mine), Dean and Donna Otteson Mineralogist, 76, 602–610. (Royston Mine), Virgil Leuth (NM Tech), Phil Weigand (Museum of Northern Prost, R., Dameme, A., Huard, E., Driard, J., and Leydecker, J.P. (1989) Infrared Arizona), and Anton Chakhmouradian (University of Manitoba) for the samples study of structural OH in kaolinite, dickite, nacrite, and poorly crystalline used in this study. This work was funded by Canada Research Chairs in Crystal- kaolinite at 5 to 600 K. Clays and Clay Minerals, 37, 464–468. lography and Mineralogy to FCH and Environmental and Isotope Geochemistry Rancourt, D.G. (1994) Mössbauer spectroscopy of minerals. I. Inadequacy of to M.F., by a Major Facilities Access Grant, Research Tools and Equipment, and Lorentzian-line doublets in fitting spectra arising from quadrupole splitting Discovery Grants to F.C.H. and M.F. from the Natural Sciences and Engineering distributions. Physics and Chemistry of Minerals, 21, 244–249. Research Council of Canada, by Canada Foundation for Innovation Grants to F.C.H. Rancourt, D.G. and Ping, J.Y. (1991) Voigt-based methods for arbitrary-shape and M.F., and by the National Science Foundation under Grant No. 0609638. Any static hyperfine parameter distributions in Mössbauer spectroscopy. Nuclear opinions, findings, and conclusions or recommendations expressed in this material Instruments and Methods in Physics Research, B, 58, 85–97. are those of the author(s) and do not necessarily reflect the views of the National Reddy, B.J., Frost, R.L., Weier, M.L., and Martens, W.N. (2006) Ultraviolet-visible, Science Foundation. near infrared and mid infrared reflectance spectroscopy of turquoise. Journal of Near Infrared Spectroscopy, 14, 241–250. referenCes Cited Robinson, K., Gibbs, G.V., and Ribbe, P.H. (1971) Quadratic elongation: a Adler, H.H. and Kerr, P.F. (1963) Infrared absorption frequency trends for anhydrous quantitative measure of distortion in coordination polyhedral. Science, 172, normal carbonates. American Mineralogist, 48, 124–137. 567–570. Amthauer, G. and Rossman, G.R. (1984) Mixed valence of iron in minerals with Sherman, D.M. (1985) SCF-Xα-SW MO study of Fe–O and Fe–OH chemical cation clusters. Physics and Chemistry of Minerals, 11, 37–51. bonds; Applications to the Mössbauer spectra and Magnetochemistry of 3+ Belyaev, A.A. and Ievlev, A.A. (1989) Arctic turquoise of Pai-Khoy (USSR). Polar hydroxyl-bearing Fe oxides and silicates. Physics and Chemistry of Miner- Research, 7, 149–151. als, 12, 311–314. Cid-Dresdner, H. (1965) Determination and refinement of the crystal structure Sklavounos, S., Ericsson, T., Filippidis, A., Michailidis, K., and Kougoulis, C. (1992) Chemical, X-ray and Mössbauer investigation of a turquoise from the of turquoise, CuAl6(PO4)4(OH)8.4H2O. Zeitschrift für Kristallographie, 121, 87–113. Vathi area volcanic rocks, Macedonia, Greece. Neues Jahrbuch für Mineralogie Cid-Dresdner, H. and Villarroel, H.S. (1972) Crystallographic study of rashleighite, Monatshefte, 469–480. a member of the turquoise group. American Mineralogist, 57, 1681–1691. Van Wambeke, L. (1971) The problem of cation defiency in some phosphates due Da Costa, G.M., Karfunkel, R.S.J, Bermanec, V., and de Sá Carneiro Chaves, M.L. to alteration processes. American Mineralogist, 56, 1366–1384. (2005) 57Fe-Mössbauer spectroscopy on natural eosphorite-childrenite-ernstite samples. Physics and Chemistry of Minerals, 31, 714–720. De Resende, V.G., Da Costa, G.M., De Grave, E., and Van Alboom, A. (2008) Möss- bauer spectroscopic study of synthetic leucophosphite, KFe2(PO4)2(OH)·2H2O. American Mineralogist, 93, 483–487. Fleet, M.E. (1976) Distortion parameters for coordination polyhedra. Mineralogical Magazine, 40, 531–533. Foord, E.E. and Taggart, J.E. Jr. (1998) A reexamination of the turquoise group: the mineral aheylite, Planerite (redefined), turquoise and coeruleolactite. Manuscript received July 23, 2010 Mineralogical Magazine, 62, 93–111. Manuscript accepted May 13, 2011 Giuseppetti, G., Mazzi, F., and Tadini, C. (1989) The crystal structure of Manuscript handled by M. darby dyar